Geology Reference
In-Depth Information
automated algorithm for lead detection and ice type classi-
fication within the leads. They used high‐resolution air-
borne visible imagery acquired by the Digital Mapping
System (DMS) with a spatial resolution between 0.1 and
2.7 m, depending on the aircraft altitude. The DMS is the
highest resolution (10 cm) visible airborne imagery system
that has flown over sea ice. The data were collected during
NASA's Operation Ice Bridge mission. The algorithm
detects a wide variety of leads ranging from narrow meter‐
wide to wide leads of hundreds of meters in width. The
unique feature of this algorithm is its ability to account for
the variation of optical signature of the leads. Although the
signature of a lead can generally be of low‐intensity and
low‐frequency distribution (atypical of the OW signature),
it may have a dynamic pixel intensity range that varies
across individual scenes. The range accounts for the differ-
ent phases of refreezing within the lead as well as the frost
flowers that may cover its thin ice surface temporarily (sec-
tion 9.4). The algorithm starts with a transformation, called
lead vicinity transformation (LVT), to determine if leads
exist in the image:
a down‐looking passive microwave radiometer, both
mounted on the underside of the NSF/NCAR C‐130 air-
craft. Data were obtained from the vicinity of the SHEBA
site during April-July 1998. The authors used an auto-
mated methodology to determine lead fraction, width, and
orientation. They found that leads were mostly covered
with thin ice. Width varied between 25 m and nearly half a
kilometer, but leads narrower than 100 m accounted for
75% of the width distribution. Another interesting finding
was that the lead fraction (percentage of lead area per unit
area of ice cover) decreases exponentially with increasing
lead width. The study confirmed, once again, that lead ori-
entation was consistent with prevailing wind patterns and
ocean circulation. Detailed survey of leads using airborne
cameras is the best tool for geometric characterization of
leads across the vast pack ice in the Arctic.
As mentioned before, the interest in lead detection has
been further aroused by the need to determine ice free-
board that allows retrieval of ice thickness from altime-
ter data. The reference water surface has to be measured
from available leads within the altimeter's coverage. The
laser altimeter system (also known commercially as
lidar) onboard ICESat‐1 was used to map sea ice thick-
ness and for that matter detect leads. For more informa-
tion about the system, the method of ice thickness
retrieval, and the importance of lead detection for this
purpose see section 10.4. Unambiguous identification of
open leads from the laser altimeter is based on combin-
ing the waveform characteristics of the return laser pulse
with the reflectivity and elevation change. This is
explained in
Kwok et al
. [2004] as well as
Zwally et al
.
[2008]. The latter reference also presents a technique for
fusion of visible imagery with altimetery data for identi-
fying leads and retrieving sea ice freeboard. A striking
illustration of using ICESat‐1 profiles of reflectivity and
elevation to identify leads imaged in near‐coincident
Radarsa‐1 image of Arctic ice is presented in
Kwok et al
.
[2007]. Both reflectivity and elevation from a lead are
much lower than the surrounding ice pixels. The status
of the lead as to whether it is open or refrozen can also
be identified from the reflectivity of the laser altimeter
return.
LVT( ,)a
Ixy
(
)
tan( (
I xy
,
))
0
,
(9.13)
2
where
I
(
x
,
y
) is the pixel intensity at the pixel position
(
x
,
y
). The transformation acts to smooth
I
over ice floes
and creates an intensity transition within the lead. This
step is followed by another affine time‐frequency trans-
formation that generates localization around low‐inten-
sity and low‐frequency lead pixels. It is called the minimal
signal transform (MST) [
Onana et al
., 2013]. An example
of the results from the two transformations is presented
in Figure 9.10. The original image from the DMS air-
borne system is shown along with the results from apply-
ing the LVT followed by the MST. The original image
shows a variety of signatures within the lead and the sur-
rounding ice. The result from the application of the LVT
transformation only (not shown in the figure) reveals less
variability of the lead signature and therefore allows eas-
ier detection. The application of the LVT followed by
the MST highlights the lead area with more uniform and
homogeneous signatures and better contrast with the sur-
rounding ice. An appropriate threshold can then be used
to discriminate between the lead and its surrounding ice
cover. The ice type classification within the lead is also
shown in Figure 9.10. The classification is based on a set
of thresholds that was determined from the probability
distribution of the fusion of minimal signal. Ice is visible
at the boundaries of the lead.
High‐resolution data needed to identify narrow leads
(a few tens of meters width) can be obtained from airborne
sensors or a space‐borne laser altimeter.
Tschudi et al
.
[2002] used data from an airborne video camera and
9.3. Surface meLt
Ice surface melt starts in late spring and extends over
the summer months (June to September in the Arctic
region). Surface temperature and albedo are the prime
indicators of surface melt but microwave data (emission
and radar backscatter) can also provide clues. Operational
ice centers are interested in surface melt because it
represents the first stage of the ice decay with its favorable
reduction in ice loads. The latter enhances the opportunity
of marine navigation through sea ice, particularly in the
